Sony Developing Gigapixel Satellite Imaging

holy_calamity writes "Sony and the University of Alabama are working on a gigapixel resolution camera for improved satellite surveillance. It can see 10-km-square from an altitude of 7.5 kilometres with a resolution better than 50 centimetres per pixel. As well as removing annoying artefacts created by tiling images in Google Earth and similar, it should allow CCTV surveillance of entire cities with one camera. 'The trick is to build an array of light sensitive chips that each record small parts of a larger image and place them at the focal plane of a large multiple-lens system. The camera would have gigapixel resolution, and able to record images at a rate of 4 frames per second. The team suggests that such a camera mounted on an aircraft could provide images of a large city by itself. This would even allow individual vehicles to be monitored without any danger of losing them as they move from one ground level CCTV system to another.'"

granted, the article got the title wrong

[everphilski]

but the school is the University of Alabama in Huntsville [uah.edu]. w00t!

Right State and System, Wrong University

[johnalex]

As much as I'd like to claim credit for my alma mater and this project, the authors didn't check the facts thoroughly. The university involved is the University of Alabama in Huntsville, not The University of Alabama. The University of Alabama [ua.edu] is located in Tuscaloosa, Alabama and boasts its own ranked engineering programs.

Let's give the Huntsville program its due.

Re:7.5 km?

[Tsu-na-mi]

That was the first thing that stuck out to me as well. In the actual application, it references an example where the camera, mounted on an airplane, flying at a height of 7.5km, can do X. The writer at New Scientist should have been clearer. Obviously, satellites do not fly at 7.5km altitude -- 75km maybe.

Re:Transmitting that much data

[kaiser423]

There are already a number of satellites doing hundreds of megabytes a second down-links. You just need a big, sensitive dish on the ground, and a good-sized transmitter. Heck, with XM and Sirius satellites with a 7 meter dish I can easily see 70dB S/N ratios without even pointing it at the satellite. Since you need about 14dB SNR to pass a couple megabytes a second pretty error-free, a signal 56dB (~400,000 time stronger) above that should be able to pass obscene amounts of data. That part has been done before.

Not impressed.

[Entropius]

A Canon 1DS Mark 3, the current speed-demon, comes close. It's got less resolution (I think they're somewhere around 20 MP) and about the same framerate (if you can get the data off the camera onto an SD card fast enough, but rigging up a custom data readout for a satellite isn't that hard.) From the specs quoted in the article (15km square from a height of 7.5km), they're using a seriously wide-angle lens setup on this thing. Sticking a tele lens (70-100mm, probably) on the Canon will probably give you about the same meters/pixel resolution, at the cost of a narrower field of view. Now just mount two of them on the satellite, if you insist on the same level of performance as the one in the article. (That'll give you about 50 mpix/sec; you can have that spread over whatever field of view you want by choice of lens.)

Also, consumer cameras (if you can call the 1Ds that, they're $thousands) have these nice things called zoom lenses. Just mount 70-200 IS zooms on the thing, and you can blow up anything you want even more detail on, at the cost of some resolution. You get the added benefit of not caring about vibration isolation on an airplane, since it's built into the lens.

Note that the only reason to use such expensive hardware is speed and a lack of complexity; a larger array of cheaper cameras would do just as well. My $400 consumer camera (Panasonic FZ50) can resolve 18 cm/pixel from 7.5 km. Hell, for the weight these things might be better than the Canons (they're much lighter); just mount however many of them you want on a plane and go. (Granted, you'd need a lot of them; they don't push out high-res images that fast.)

There's no reason to use custom-built hardware when Canon (or Nikon or Panasonic or whoever) is already mass-producing stuff that will get the job done cheaper, with more flexibility (zoom lenses, ability to add more cameras or swap lenses.)